3
$\begingroup$

Suppose T is a bounded operator on a Hilbert space. Show that if λ is in the residual spectrum of T, then $\bar{λ}$ is in the point spectrum of the adjoint.

Here is what I think needs to be done. We know that $\langle Tu,v\rangle = \langle u,T^*v\rangle = \overline{\langle T^*v,u\rangle}$. Does that help with connecting $\lambda$ with $\overline{\lambda}$?

$\endgroup$
  • $\begingroup$ < and > mean "less than" and "greater than". The correct symbols you mean are \langle and \rangle. $\endgroup$ – Zev Chonoles Mar 17 '15 at 5:03
  • $\begingroup$ Okay, I will edit the question. $\endgroup$ – Bobby Jones Mar 17 '15 at 5:25
  • 3
    $\begingroup$ Maybe can help the fact $\overline{R(T-\lambda I)}=N(T^*-\overline{\lambda}I)^{\bot}$. Since $\lambda$ is in the residual spectrum of $T$, then $\overline{R(T-\lambda I)}\neq H$. And then exist an element $x\neq 0$ in $N(T^*-\overline{\lambda} I)$. $\endgroup$ – MrSelberg Mar 17 '15 at 6:12
  • $\begingroup$ hm where does that fact come from? and what is R&N? $\endgroup$ – Bobby Jones Mar 17 '15 at 6:40
  • 1
    $\begingroup$ As Hee Kwon Lee pointed, $R$ is the range of the operator, $N$ is the Kernel. For a proof of this fact, see Functional Analysis, Rudin, Theorem 4.12 (page 94). $\endgroup$ – MrSelberg Mar 17 '15 at 7:02
1
$\begingroup$

I will explain Selberg's solution : Since $R:=\overline{R(T-\lambda I)}$, i.e., range of $T-\lambda I$, is a closed subspace and $H\neq R$ (since $\lambda$ is a residual spectrum), then there exists $$ x_0\neq 0 \in V,\ H=R\oplus V $$

So $$ 0= \langle (T-\lambda I)x,x_0\rangle=\langle x, (T^\ast -\overline{\lambda } I ) x_0 \rangle $$

Since $x$ is arbitrary, $(T^\ast -\overline{\lambda } I ) x_0 =0$.

$\endgroup$
  • $\begingroup$ I see. Can someone explain why we use the kernel? Also, can someone suggest a good resource for learning about spectrum? $\endgroup$ – Bobby Jones Mar 17 '15 at 7:05
  • 1
    $\begingroup$ @BobbyJones We want to see the kernel because we want to prove something about it: we want to prove that the kernel is not trivial. Hence I thought in link the range of $T-\lambda I$ (the information we know) and the Kernel of $T^*-\bar{\lambda}I$. (bad english :( ) $\endgroup$ – MrSelberg Mar 17 '15 at 7:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.